[ECOS] ARM startup stack & C++

[ECOS] ARM startup stack & C++

[Robin Farine]

Hi gang,

I have got some stack overflow during execution of non-trivial static
initializers with C++ programs. The file vectors.S for ARM declares a startup
stack of 512 bytes when configured for separate interrupt stack. What do you
think I should do to get a big stack for initializers/finalizers?

1) Change the hardcoded 512 into 8192 ;-)

2) Add a configuration option for the startup stack size but nothing will use
   this "big" area afterwards, correct?

3) Disable the separate interrupt stack but this probably has annoying
   implications.

4) A completely different and wonderful solution?


Number 2 seems not too bad to me, what's your opinion?

Robin

Re: [ECOS] ARM startup stack & C++

[Andrew Lunn]

On Wed, Aug 08, 2001 at 05:15:31PM +0200, Robin Farine wrote:
> Hi gang,
> 
> I have got some stack overflow during execution of non-trivial static
> initializers with C++ programs. The file vectors.S for ARM declares a startup
> stack of 512 bytes when configured for separate interrupt stack. What do you
> think I should do to get a big stack for initializers/finalizers?
> 
> 1) Change the hardcoded 512 into 8192 ;-)

Not a good idea. The patch will get lost/forgotten when you hand the
completed code over the industrilisation team of intergration. 

> 
> 2) Add a configuration option for the startup stack size but nothing will use
>    this "big" area afterwards, correct?

Sounds reasonable. Anyone with a non-trivial static initialiser
probably has enough MB of memory that 8K is will not be missed. 

> 4) A completely different and wonderful solution?

Change your static to dynamic initialisationm so it can be done in the
context of a thread you can control the size of the stack.

        Andrew

Re: [ECOS] ARM startup stack & C++

[Jonathan Larmour]

Andrew Lunn wrote:
> > 4) A completely different and wonderful solution?
> 
> Change your static to dynamic initialisationm so it can be done in the
> context of a thread you can control the size of the stack.

5) Enable CYGSEM_LIBC_INVOKE_DEFAULT_STATIC_CONSTRUCTORS and then they are
run from the context of the thread that runs main().

Jifl
-- 
Red Hat, Rustat House, Clifton Road, Cambridge, UK. Tel: +44 (1223) 271062
Maybe this world is another planet's Hell -Aldous Huxley || Opinions==mine

Re: [ECOS] ARM startup stack & C++

[Robin Farine]

Jonathan Larmour <jlarmour@redhat.com> writes:

[...]

> 5) Enable CYGSEM_LIBC_INVOKE_DEFAULT_STATIC_CONSTRUCTORS and then they are
> run from the context of the thread that runs main().

Et voila, so simple that I can't believe it :-)

Thanks Jonathan,

Robin

Re: [ECOS] ARM startup stack & C++

[Robin Farine]

Andrew Lunn <andrew.lunn@ascom.ch> writes:

[...]

> 
> Sounds reasonable. Anyone with a non-trivial static initialiser
> probably has enough MB of memory that 8K is will not be missed. 

Agreed.

> > 4) A completely different and wonderful solution?
> 
> Change your static to dynamic initialisationm so it can be done in the
> context of a thread you can control the size of the stack.

I'm already doing something like this but even the C++ runtime support pulled
from libgcc causes a stack overflow.

Thanks,

Robin